A Novel Machine Learning–Based Approach for Characterising the Micromechanical Properties of Food Material During Drying.

Plant-based food materials (PBFMs) such as fruits and vegetables contain various irregular cellular compartments. Like other engineering materials, the characterisation of micromechanical properties (MMPs) of PBFMs is intensely important for accurately estimating the functionality of dried food products. The application of a machine learning (ML)–based approach to characterise the MMPs is a promising idea. However, no intensive research in this regard has been attempted yet. Therefore, we proposed an ML-based modelling framework to characterise the MMPs of PBFMs during drying. A feed-forward artificial neural network (ANN) model with a backpropagation algorithm was developed and optimised with a genetic algorithm (GA)–based optimisation tool for characterising PBFMs, specifically carrots. Moreover, the accuracy of the ANN model was compared with a multiple nonlinear regression (MNLR) model. It was found that the developed network model agreed very well with the experimental data when predicting the elastic modulus, stiffness and hardness, with an accuracy of the goodness of fit (R2) values of 0.992, 0.993 and 0.802, respectively. It is expected that the developed model has incredible potential to characterise the MMPs of similar food products. [ABSTRACT FROM AUTHOR]